Modelpublic

theprint/Genuine-Zeth-4B-GGUF

sourceHugging Faceapache-2.0updated 1y agoView on Hugging Face

0likes73downloads

Model Card

Genuine-Zeth-4B - GGUF Quantized

Quantized GGUF versions of Genuine-Zeth-4B for use with llama.cpp and other GGUF-compatible inference engines.

Original Model

—Base model: theprint/Zeth-Gemma3-4B
—Fine-tuned model: theprint/Genuine-Zeth-4B
—Quantized by: theprint

Available Quantizations

—Genuine-Zeth-4B-f16.gguf (8688.3 MB) - 16-bit float (original precision, largest file)
—Genuine-Zeth-4B-q3_k_m.gguf (2276.3 MB) - 3-bit quantization (medium quality)
—Genuine-Zeth-4B-q4_k_m.gguf (2734.6 MB) - 4-bit quantization (medium, recommended for most use cases)
—Genuine-Zeth-4B-q5_k_m.gguf (3138.7 MB) - 5-bit quantization (medium, good quality)
—Genuine-Zeth-4B-q6_k.gguf (3568.1 MB) - 6-bit quantization (high quality)
—Genuine-Zeth-4B-q8_0.gguf (4619.2 MB) - 8-bit quantization (very high quality)

Usage

With llama.cpp

bash

# Download recommended quantization
wget https://huggingface.co/theprint/Genuine-Zeth-4B-GGUF/resolve/main/Genuine-Zeth-4B-q4_k_m.gguf

# Run inference
./llama.cpp/main -m Genuine-Zeth-4B-q4_k_m.gguf \
  -p "Your prompt here" \
  -n 256 \
  --temp 0.7 \
  --top-p 0.9

With other GGUF tools

These files are compatible with:

—llama.cpp
—Ollama (import as custom model)
—KoboldCpp
—text-generation-webui

Quantization Info

Recommended: q4_k_m provides the best balance of size, speed, and quality for most use cases.

For maximum quality: Use q8_0 or f16 For maximum speed/smallest size: Use q3_k_m or q4_k_s

License

apache-2.0

Citation

bibtex

@misc{genuine_zeth_4b_gguf,
  title={Genuine-Zeth-4B GGUF Quantized Models},
  author={theprint},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/theprint/Genuine-Zeth-4B-GGUF}
}